►
From YouTube: NUG Monthly Webinar, December 15, 2022
Description
NUG Monthly Webinar on December 15, 2022. The topic of the day is Allocation Year Transition.
A
So
here's
sort
of
our
schedule
for
today
so
first
of
all,
this
should
be
an
interactive
meeting.
I,
don't
wanna
I'm
here
and
waiting
for
you
all
to
say
anything.
So,
please
participate!
You
can
raise
your
hand
or
you
can
just
speak
up
since
we
only
have
nine
people
to
speak
up.
I
think
that's
the
best
best
way
to
do
it
and
then,
of
course,
you
can
also
join
in
discussions
in
the
webinars
channel
on
the
nurse
user
slack
while
we're
having
this
meeting
okay.
A
So
today's
agenda
first
we're
going
to
talk
about
winner
of
the
month
and
then
today,
I
learned
we're
gonna,
have
some
announcements
and
cmps
and
then
we're
gonna
have
the
topic
of
the
day,
which
is
the
allocation
year
transition
and
my
colleague
Helen.
He
is
here
to
talk
to
you
all
about,
what's
happening
for
that
and
then
we'll
talk
about
you
know
if
you
all
have
any
suggestions
for
future
topics
of
what
you
might
be
interested
in
hearing
about
and
finally,
we'll
do
last
month's
numbers.
A
So
let's
get
started
with
the
win
of
the
month.
So
this
is
your
time
to
shine.
Do
you
have
an
achievement
that
you
want
to
share,
or
you
have
a
shout
out
to
someone
else's
achievement
like
a
paper
accepted
Soft
bug,
science
achievement
an
Innovative
use
of
high
performance
Computing.
Anybody
want
to
share
anything
that
you've
accomplished
this
month
or
someone
that
you
know
has
accomplished
this
month.
B
One
is
the
it
is
submitted,
not
accepted.
Yet
there
are
some
papers
every
that
describes
technical
aspects
of
the
climate
simulations.
We
run
on
mainly
on
Quarry,
some
of
Edison
some
interesting
aspects
of
this
paper
being
again
technical.
We
for
first
we
know
that
our
computer
code,
I
mean
model
code,
is
around
slowly
on
Canal,
compared
to
Haswell
or
Edison,
so
I
really
digged
into
some
literature,
published
I,
actually
asked
staff
and
and
ASAP.
B
You
know
some
studies
actually
dig
into
part
of
the
CSM
code
we
used
and
then
why
it
is
slow,
basically,
lack
of
open,
MP
and
the
lack
of
vectorization.
It
doesn't
really
make
compiler
easy
to
vectorize
and
then
there's
some
other
among
some
other
aspects.
So
we
did
pointed
it
out
and
cited
those
papers
in
in
our
paper,
so
that
the
people
being
aware
how
those
Legacy
full-time
code
does
not
fully
taking
advantage
of
this
recent
many
core
architecture
in
parameter.
B
Cpu
node,
is
also
many
core
architecture,
even
though
it
has
more
memory,
so
probably
kind
of
model
called
skill
rounds
faster
than
k,
l.
So
anyway,
that's
one
accomplishment.
The
other
is
I
am
also
contributing
this
new
special
interest
group
at
nask
for
another
current
model
called
wolf,
so
we
started
so
we
use
a
group
to
enhance
communication
among
the
users,
Running,
Wolf
and
nurse,
and
then
to
explain
or
to
even
contribute
documentation
on
how
to
compile
and
the
best
practice
on
how
to
run
this
model.
B
And
then
we
apply
for
allocation,
and
then
we
just
got
allocation
for
the
rest
of
this
year
and
we're
gonna
follow
up
with
the
renewal
allocation
for
the
next
year
and
we
started
the
benchmarking
simulation
on
Palmetto
and
also
we
are
drafting
the
narsk
documentation.
So
hopefully
this
gets
reviewed
sometime
early
next
two
years,
so
that
all
the
wolf
model
users
are
ready
to
run
the
model
on
Parramatta
when
the
transition
happens.
So
those
two
are
my
accomplishments.
A
C
A
A
B
You
again
sorry
to
really
I,
don't
know
occupying
the
stage
this
time,
but
I
think
good
things.
I
learned
is
the
a
few
applications
I've
been
using
for
my
workflow,
pathetic
analyzing
those
data
on
a
net
CDA
file
like
NCO,
NCL
and
NC
View
at
first
I
thought,
wow,
there's
no
modules
on
power,
mother,
so
I
thought
I
have
to
install
myself,
but
once
I
load,
spec
and
e4s
start
module,
then
I
can
actually
I
can
find
all
those
are
installed
in.
B
In
the
the
at
least
newest
E4
s20
version
to
do
on5,
so
I
recommend.
When
we
go
through
the
power
model
we
didn't
find.
If
we
don't
find
the
applications
we
used
to
use
Maybe
still,
you
know
at
least
try
lording
back
and
there
was
a
loading
fos
to
check
and
if
not
I
have
some
I
think
one
Library
I
don't
find
so
I
might
try
to
use
fact
to
install
that
for
myself
and
I
might
report
that
later
here.
B
B
A
C
E
That's
that's
great
feedback.
You
know
over
the
years
we've
kind
of
gone
back
and
forth
on
whether
this
should
be
group
or
Project,
Director
or
directory,
or
individually
based
quotas
and
sometimes
we're
limited
by
the
technology
we
happen
to
have,
but
but
you're
right
we
should.
We
should
mention
that
very
high
in
the
documentation,
yeah.
A
Yeah
definitely
and
then
your
first
advice
about
about
SPAC
or
sorry
e4s,
and
then
it's
back
load
all
these
different
libraries
and
applications
and
things
that's
absolutely
correct.
They
have
like
I,
don't
know
several
hundred
different
different
things
that
you
can
learn
from
that
e4s
module.
So
it's
it's
a
pretty
amazing
Treasure
Trove.
If
you
kind
of
look
through
it
and
see
what's
there
for
you,
so.
A
Sure,
okay,
well
I'll,
share
something
that
I
learned
within
the
past
month.
So
you
know
there's
this
quote
right
down
here.
If
we
knew
what
we,
what
it
was
we
were
doing,
it
would
not
be
called
research
with
it
and
it's
attributed
to
Einstein
somebody.
Somebody
told
me,
you
know
that's
apocryphal.
At
best,
like
yeah
I
figured,
it
was
not
really
true,
but
anyway
the
point
of
it
still
stands.
A
I'm
actually
looking
the
the
nug
slack
I
think
in
the
general
channel
for
for
that
discussion
from
last
month,
and
it
was.
He
also
included
a
link
to
a
really
interesting
article
about
attributions
in
papers
and
how
things
kind
of
can
become
kind
of
like
a
game
of
telephone.
Where
you
know
somebody
says
that
the
first
author
said
this
thing,
but
that's
not
exactly
right,
and
then
the
next
person
quotes
that
a
second
paper
while
saying
that
the
first
author
said
something
and
it
just
becomes
more
and
more
distorted
anyway.
A
A
A
C
A
Well,
announcements
and
see
a
piece,
so
let's
talk
about
there's
two:
a
really
good
cause
for
participation,
both
from
the
from
the
the
Krill
Institute.
So
the
first
one
is
the
the
James
Coronas
award
and
Leadership
community
building
and
communication.
So
there's
there's
a
listing
nominations
for
that.
The
nominations
are
due.
December
31st
I
happen
to
know
that
this
is
a
great
word,
because
I
wouldn't
I
was
the
first
winner
of
this
award.
A
So
if
you,
if
you
know
somebody
who's,
a
mid-career
scientist
who
is
really
a
big
leader,
Community
Builder
and
Communicator
then
be
sure
to
nominate
them
for
that
award
and
I'm.
Also
on
the
award
evaluation
committee
nowadays-
and
it's
it's
really
great
and
you
get
to
see
all
these
really
amazing
people
doing
amazing
work.
Okay,
so
then
the
second
one
is
the
doe
computational
science
graduate
Fellowship.
A
A
We've
got
one
upcoming
event
currently
scheduled
it's
gonna,
be
in
January.
We
were
so
close
to
the
holidays,
nothing's
happening
right
now,
and
it's
the
ideas,
ECP
webinar,
and
it's
about
something
called
openscapes
supporting
better
science
for
future
us.
It
has
to
do
with
open
science
and
and
keeping
science
open
anyway.
A
Very
it
should
be
a
really
interesting
webinar
and
I'm
gonna
post
these
slides
after
the
web
after
this
session
and
so
you'll
be
able
to
click
on
those
links
and
then
finally,
I
just
want
to
really
stress
the
Corey
to
promoter
transition
office
hours.
So
we've
held
five
office
hours
since
November,
and
we've
met
with
more
than
90
users
and
I.
Think
it's
been
really
helpful
for
people.
A
Pacific
time
you
can
come
see
us
in
our
core
to
promoter
transition
office
hours
and
then,
if
you,
if
you
can't
make
that
one
we'll
have
another
one
on
Friday,
January,
6th
and
another
one
on
Thursday,
January,
12th
and
so
I
think
this
is
a
really
great
opportunity
for
people
and
we
have
experts
from
all
of
our
different
user-facing
groups.
They'll,
be
on
hand
there
and
so
they'll
be
able
to
help
you
with
a
wide
variety
of
of
issues.
A
G
C
A
A
lot
of
positions
open
right
now.
This
is
not
even
a
complete
list.
I
just
put
the
ones
that
I
thought
would
probably
be
most
appealing
to
this
audience,
but
so
there's
several
new
ones.
So
the
HPC
user
environment,
architect
that
one
just
came
out
this
week
and
then
the
science
engagement
engineer
and
HPC
consultant.
They
were
new
this
past
week
and
then
all
of
the
others
that
you
can
see
there
so
yeah
we're
actively
hiring
so
we'd,
be
really
interested.
A
H
Next
slide,
please
so
I
will
cover
these
topics:
the
New
Year's
allocations
and
the
transition
process.
What
happens
on
the
start
day
of
the
due
allocation
and
what
will
be
changing
in
the
new
allocation
year?
What
about
the
discontinued
users
next
slide?
Please,
the
allocation
of
2023
is
from
January
18th
Wednesday
to
Tuesday
January
16th
of
the
next
year.
H
H
The
next
year's
allocation
capacity
is
all
based
on
Perimeter
hours,
but
the
allocated
CPU
hours
can
be
used
for
query
and
for
model
CPU.
The
allocated
GPU
allocation
will
be
used
for
a
per
model
GPU
when
we
were
charged
next
slide,
yeah.
So
sorry
we're
Church
starting
from
second
day
the
Thursday
of
January
19th
next
slide.
Please.
H
First
of
all,
the
IRS
database
needs
to
be
updated
with
the
new
allocation
data,
such
as
projects
awarded
users,
your
CPU
and
GPU
allocations
and
HPS
storage
allocation.
Awards
then,
on
each
of
the
computational
systems,
users
and
projects
data
needs
to
be
synced
up
with
Iris
database
about
active
users,
and
we
also
will
need
to
on
this
at
the
transition.
We
also
need
to
clean
up
batch
jobs
that
do
not
have
the
new
allocations
or
users
will
sometimes
have
system
Engineers
happening
on
the
on
the
transition
day,
just
to
court
to
to
coordinating
those.
H
H
So
shortly
before,
ay
2022,
the
sum
of
the
for
the
transition
already
started.
For
example
the
week
before
the
AY
start,
we
will
have
no
new
user
account
creation
and
validation,
and
since
October
2022,
we
are
not
accepting
ercap
requests
for
new
projects,
then
starting
from
the
one
allocation
email
award
email
is
sent
out.
We
will
open
up
a
iris
staging
API
for
the
pis
to
do
something
to
prepare
for
next
year,
especially
for
the
continued
projects.
H
You
have
the
deadline
from
now
until
when
we
cash
before
a
new
ay,
that
pis
must
nominate
which
users
will
continue.
You
have
the
API
e-logging,
and
do
you
see
that
that
the
other
users
are
not
automatically
renewed
into
the
new
allocation
year?
The
pis
have
to
must
have
to
check
those
boxes
too,
for
the
new
continued
users
and
also
pis
have
to
check
and
update
which
users
will
have
the
premium
qos
access.
Just
for
your
convenience.
H
We
have
inherited
the
AY
22s
allowed
list
of
Premium,
but
you
need
to
check
and
update
that
also
before
the
deadline,
and
we
have
an
instruction
link
for
you
to
follow
how
to
do
those
there's.
Also,
like
there's
a
toggle
that
you
can
allow
all
users
to
continue,
you
can
also
allow
all
users
to
have
premium
just
for
your
convenience.
There's
a
toggle
buttons,
just
a
reminder
that
before
the
end
of
this
year,
it's
it's
great.
H
It's
a
good
idea
for
pis
or
users
to
check
all
your
premium
jobs
in
the
queue
right
now,
if
you
don't
do
anything,
and
if
those
jobs
happen
to
start
in
ay
23,
you
might
have
some
unexpected
bigger
charges
and
also,
if
you
forgot
to
renew
a
user
that
has
premium
jobs
in
the
queue,
then,
when
the
new
year
start,
those
jobs
will
be
deleted
because
those
users
don't
have
the
premium
access,
so
double
check
your
jobs
in
the
queue
shortly
before
the
new
ay
starts
next
slide.
Please.
H
H
H
The
plan
is
to
have
no
downtime,
and
but
there
will
be
some
like
there
will
be
live
updating
from
IRS,
so
you
might
expecting
some
slurm
command
slowness
during
the
sinking
with
Iris,
but
it's
very
short
very
transient,
and
it
happens
at
different
time,
maybe
on
different
login
nodes,
Etc
and
expecting
all
other
systems
and
services
up
at
this
point
next
slide.
Please
there'll
also
be
a
lot
of
changes
for
the
jobs
in
the
queue.
Basically,
all
the
below
jobs
will
be
deleted
on
both
query
and
parameter
so
for
jobs.
H
If
you,
the
project
is
not
continuing
all
the
jobs
that
search
it
with
that
project
will
be
deleted
if
a
user
is
not
continuing
in
a
project
that
user's
job
will
be
deleted.
If
a
user
has
no
access
to
the
premium
and
has
a
premium
jobs
in
it,
those
jobs
will
be
deleted.
All
the
overrun
jobs
will
be
deleted
because
at
start
of
the
new
allocation
you
would
have
Pro
have
hours
and
also,
according
to
our
policy,
our
jobs
being
held
older
than
12
weeks
will
be
deleted
next
slide.
Please.
H
Then
after
ay
2023
starts,
we
would
like
to
remind
the
pis
you
should
go
into
the
iris
database
and
check
the
each
of
your
users,
the
CPU
and
TPU
allocation
hours
or
percentage
that
you
have
set
for
them
for
your
convenience.
We
have
inherited
the
our
percentage
allowed
from
last
year,
but
you
might
want
to
change,
and
especially
for
for
the
GPU
allocations,
because
there
was
there
were
new
starting
this
year.
H
H
So
one
big
change
for
this
year's
policy
is
that
newscast
added
modified
our
appropriate
user
policy
with
the
addition
of
a
user
code
of
conduct.
This
is
designed
to
express
establish
our
user
Community
norms
for
how
we,
our
community
interact
with
one
another.
The
community
includes
nurse
staff,
our
about
9000
users,
vendors
speakers,
Etc,
the
the
club
we
would
like
to
Foster
and
future
enhancing
our
collaborative
culture
of
respect,
fairness
and
inclusion.
H
Lots
of
examples
of
communication
happens
with
emails,
slack
messages,
user
tickets,
training
events,
documentations
conferences,
working
groups,
some
clubs,
meetups
all
sorts
of
communications.
We
would
like
to
establish
this
community
Norm
that
we
have
this
user
code
of
conduct,
there's
more
information
in
the
FAQ
page
and
October,
not
talk
by
Rebecca.
H
H
You
will
be
able
to
do
so
once
the
AY
starts
after
January
18th,
when
you
log
into
iOS
you'll,
see
an
API
pop-up
and
you
have
you
would
click
and
accept
the
terms
of
the
agreement.
So
any
user
who
does
not
complete
this
task
by
February
20th
will
be
considered
as
discontinued
users
and
their
account
will
be
deactivated
next
slide.
Please.
So,
let's
tell
you
what
happens
to
the?
Oh
sorry,
that's
another
slide.
H
H
H
As
already
mentioned,
the
AY
2023
allocations
are
based
on
parameters
capacity,
but
so
that
we
actually
was
Corey
being
on
you
have
we
have
more
Computing
Power
for
the
year
and
also
starting
November
through
January,
we
have
been
doing
lots
of
transitioning
trainings
and
office
hours
and
the
next
office
hours
tomorrow.
The
one
training
happened
on
December.
First,
we
have
slides
and
videos
available.
H
If
you
would
like
to
check
those
out
foreign
and
also
then
in
the
late
January
early
February,
we
will
announce
the
final
date
for
the
decommissioning,
the
final
date
as
T,
then
the
final
date
minus
one
week.
What
we
want
to
do
is
we
will
put
on
a
system
reservation
so
that
jobs
were
still
around
but
won't
run
across
the
boundary
of
the
decommissioning
date,
the
the
retirement
date,
the
T
stands
for
time
and
date
and
then
Island
retirement
date.
All
the
jobs
on
in
the
queue
will
be
deleted.
H
H
H
H
A
F
E
B
E
When
Cory
goes
away,
it's
just
the
machine
is
going
away,
so
all
the
projects
will
continue
and
each
project
has
a
separate
allocation
of
CPU
time
and
GPU
time.
If
you
were
using
Corey
GPU,
we
never
charged
for
that.
So
that
was
never
anything
that
came
out
of
your
allocation,
and
so
you
should
switch
over
to
using
promoter
GPU,
but
then
that
time
will
be
charged
so
I
guess
in
a
certain
sense,
a
free
resource
is,
is
going
away,
even
though
it's
not
a
real
big
one.
E
C
G
E
H
D
H
A
Okay,
well,
thank
you
Helen!
Thank
you.
This
was
very
informative,
learned
a
lot.
If
you
all
have
any
questions,
you
can
always
check
out
that
webpage
that
that
Helen
pointed
out
like
I,
said
I'll
I'll
upload
these
slides
after
after
this
session
is
over.
I
can
see
all
that.
Okay.
Well,
thank
you
alrighty!
So,
coming
up,
what
are
we
going
to
talk
about
in
future
months?
A
I,
don't
actually
know
so
next
month,
I
think
we're
going
to
talk
about
our
user
community
and
what
our
plans
are
there
for
for
our
user
Community,
but
in
February
it's
it's
up
in
the
air.
So
please
let
us
know
if
there's
anything
that
that
you're
interested
in
having
as
a
topic-
and
we
will
talk
about
it
and
as
as
usual,
of
course,
we'd
love
to
hear.
A
A
Okay.
Last
but
not
least,
last
month's
numbers,
so
query
utilization
was
95.5,
so
well
done
everybody
large
jobs,
which
is
jobs
larger
than
at
least
1024
nodes
or
larger,
made
up
55.5
of
all
the
cycles
that
were
used
last
month,
you
all
sent
in
694
new
tickets.
We
closed
748
tickets,
so
call
us
we
closed
more
than
we
opened
so,
but
we
still
have
a
backlog
of
652
tickets
that
we
are
still
working
on
so
that
concludes
our
meeting.
Does
anybody
have
any
questions
comments?
B
Rebecca
I
just
want
to
make
sure
these
office
hours
for
Parramatta
transitions.
Oh
I
forgot
to
turn
on
my
video
does.
Do
you
guys
still
plan
to
continue
I
think
you
briefly
mentioned,
but
after
January,
because
I
do
have
some
libraries
applications
I
have
to
Port,
but
I
don't
have
a
time
before
early
January
to
be
old,
so
this
office
hour
in
February
March
will
be
also
helpful.
A
E
E
Do
that,
and-
and
you
know
in
addition
to
the
office
hours,
which
are
great-
we
are
we're.
Rethinking
our
model
of
of
nisap
for
the
next
year,
and
one
thing
I
think
we'd
like
to
do
is
to
engage
with
users
who
have
who
want
to
port
or
optimize
for
gpus
their
codes
or
workflows,
but
don't
want
to
necessarily
engage
for
like
five
years
or
something
like
that,
but
have
a
shorter
engagement,
be
it
weeks
or
months,
and
something
like
that.
E
So
it
sounds
like
you
might
be
a
good
candidate
for
that
sort
of
thing.
So
you
could
send
me
or
Rebecca,
maybe
an
email
with.
If
you
have
any
idea
about
kind
of
how
long
engagement
you
think
you
might
need
and
what
your
code
is
and
that
sort
of
thing
we
could
certainly
keep
that
on
file.
And
if
we,
if
we
implement
this,
you
might
be
one
of
the
the
top
candidates
to
engage
with
yeah.
B
B
Kind
of
motor
or
some
a
python
routines
form,
but
I
do
have
some
full
fresh
climate
model,
which
is
GPU
ready,
using
open
ACC
but
not
tested
on
power
model
and
another
Photon
called
I
have
no
idea,
no
comments
but
I
want
to
post
to
change
to
openmp
to
openvp
MP
delivery
directive
stuff,
but
I
haven't
had
a
chance
to
bring
it
to
the
help
desk
because
I
don't
have
no
idea
about
this
course.
Okay,.
E
A
C
D
C
D
D
E
C
E
A
B
A
B
A
Yeah,
so
what
happens
is
there's
you
have
two
jobs
that
are
eligible
to
increase
in
priority,
so
your
other
jobs
are
still
in
the
queue
they
just
retain
the
same
priority
until
one
of
those
jobs
runs
and
then
that
the
third,
the
formerly
third
job,
is
then
eligible
to
age.
But
in
theory,
your
your
jobs
that
are
in
lower
priority
and
the
queue
could
run,
especially
if
they
are
short
jobs,
because
those
would
fit
really
well
in
what
we
call
the
backfill.
Okay.
C
A
B
C
A
And
then
after
it
does
that
and
and
it's
at
the
end
of
that
list,
then
it
goes
down
the
whole
list
and
it
says:
is
there
anything
on
this
list
that
I
could
start
right
now
that
wouldn't
interfere
with
this
schedule
that
I've
already
built
so
that
schedule
that
it's
already
built
contains
holes
in
it
and
those
holes
are
typically
short
in
time,
but
they
could
be
actually
fairly
large
in
number
of
nodes,
I
see,
so
we
so
we
actually
find
that
your
wait.
Time
is
directly
proportional
to
the
wall.
A
B
C
B
A
project
maybe
still
have
abundant,
but
I
have
I'm
running
simulations
for
each
project,
so
sometimes
I
use
low
priority
queue
for
projects
that
has
much
less
allocation
left
and
maybe
I
still
use
regular
queue
for
other
projects
and
then,
when
I,
look
at
my
queue
of
the
job,
I
have
maybe
three
regular
jobs.
For
you
know
projects
with
adequate
amount
of
allocation,
any
one
or
two
low
priority
jobs
for
the
projects
that
has
maybe
less
computing
time
and
then
so.
B
A
B
C
H
B
But
yeah,
but
based
on
my
workflow
I,
also
kind
of
rather
ask
you
guys
to
consider
low
priority
queue
on
power
Mata
as
well,
because
that's
how
sometimes
I
still
keep
something
get
going
for
project,
even
though
the
account
is
going
well
but
still
simulation.
It's
a
long
term,
not
too
high
priority,
but
it's
still
get
going
is
better,
so
I
use
a
slow
priority.
Thank
you.
So.